In Fig XXXXXXXXXXis the body of a loop. Addresses such as a(R9) are intended to be memory locations, where a is a constant, and R9 is the register that counts iterations through the loop. You may...

In Fig. 10.31 is the body of a loop. Addresses such as a(R9) are intended to be memory locations, where a is a constant, and R9 is the register that counts iterations through the loop. You may assume that each iteration of the loop accesses different locations, because R9 has a different value. Using the machine model of Example 10.12, schedule the loop of Fig. 10.31 in the following ways:

a) Keeping each iteration as tight as possible (i.e., only introduce one nop after each arithmetic operation), unroll the loop twice. Schedule the second iteration to commence at the earliest possible moment without violating the constraint that the machine can only do one load, one store, one arithmetic operation, and one branch at any clock.

b) Repeat part (a), but unroll the loop three times. Again, start each iteration as soon as you can, subject to the machine constraints.

! c) Construct fully pipelined code subject to the machine constraints. In this part, you can introduce extra nop's if needed, but you must start a new iteration every two clock ticks.

Fig. 10.31

Example 10.12